Research map / Label-efficient and robust learning / Domain adaptation and generalization
Unsupervised Image-to-Image Translation Using Domain-Specific Variational Information Bound
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use domain-specific and domain-invariant codes with a domain-specific variational information bound instead of a one-to-one shared-latent mapping | Ours · Edges↔Shoes | LPIPS distance: 0.113 -> 0.121 | Helped |
| Use domain-specific and domain-invariant codes with a domain-specific variational information bound instead of a one-to-one shared-latent mapping | Ours · Edges↔Shoes | FID score: 43.18 -> 48.36 | Hurt |
| Use domain-specific and domain-invariant codes with a domain-specific variational information bound instead of a one-to-one shared-latent mapping | Ours · Edges↔Handbags | LPIPS distance: 0.134 -> 0.129 | No clear effect |
| Use domain-specific and domain-invariant codes with a domain-specific variational information bound instead of a one-to-one shared-latent mapping | Ours · Edges↔Handbags | Fooling rate: 34.9 -> 33.2 | No clear effect |
| Add domain-specific code cycle consistency (ablation) | Ours · Edges↔Shoes | LPIPS distance: 0.095 -> 0.121 | Helped |
| Add domain-specific code cycle consistency (ablation) | Ours · Edges↔Handbags | LPIPS distance: 0.113 -> 0.129 | Helped |
| Remove domain-invariant code cycle-consistency criterion (ablation) | Ours · Image translation | image quality | Hurt |
| Neglect minimization of mutual information between target-specific information and the source domain (ablation) | Ours · Face translation | domain-specific attribute control | Hurt |
| Set beta very small, weakening the mutual-information maximization term (ablation) | Ours · Image translation | domain-specific code meaningfulness | Hurt |
| Set beta very large (ablation) | Ours · Image translation | separation of domain-specific information | Hurt |
Closest papers by meaning
- Unsupervised Image-to-Image Translation with Density Changing Regularization (NeurIPS 2022 · similarity 0.90)
- Unsupervised Image-to-Image Translation Networks (NeurIPS 2017 · similarity 0.90)
- Multi-domain image generation and translation with identifiability guarantees (ICLR 2023 · similarity 0.89)
- Image to Image Translation for Domain Adaptation (CVPR 2018 · similarity 0.88)
- Unsupervised Image-to-Image Translation With Generative Prior (CVPR 2022 · similarity 0.88)
- Towards Identifiable Unsupervised Domain Translation: A Diversified Distribution Matching Approach (ICLR 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.